The first letter.

This is the first GLX letter. I’m writing it for two readers. The first found GLX this year, probably because someone told them about it, or they saw the address on an email and typed it in. The second is reading this in thirty years and wants to know if we kept our word. This letter is mostly for the second one.

There are no results to report yet. GLX was named in 2026, so almost every number we track is still at zero. What I can give you is the rules. Everybody who says they’ll never sell means it, right up until a big enough number shows up. I know that. So the rules are written down here, with a date on them, where you can check them.

Nobody in my family ran a business. The easy path was to become a cop, and plenty of people, me included, thought I should take it. Instead I co-founded my first company, SpartX, at 22, out of a spare bedroom. We sold it four years later.

I’ve spent the twenty years since inside healthcare operations, from the field to the chief executive’s chair. What I watched happen, over and over, is that the companies doing the real work don’t get to keep what they build. Owners sell every three to five years, and every sale resets the trust. The people who were counting on the last owner start over with the next one.

GLX exists so that doesn’t happen to the companies we build. That’s why the family name is on the door. Our name goes on everything we build, and we could never buy it back.

The machine is simple on purpose. The care companies come first. They get funded for payroll and for the bad year nobody sees coming. We just lived through one of those, and the reserves we built before it are why we’re still here.

Only after the companies are funded does anything move. What’s left goes into the GLX book every quarter, on a schedule that’s written down. Nobody decides how much or when, me included. There’s no timing the market and no borrowing.

The money buys long-term securities the week it lands. We count them in shares, not dollars, because the price moves every day and the share count only moves when we add to it. And we don’t sell. The home page shows how many shares GLX has ever sold. Today it says zero. We want it to still say zero when you read this.

The book runs on four rules. They fit on a card.

Toll booths

Nobody knows which AI companies will win. Every one of them pays for power, buildings and chips, so that’s where we hold the most. Own the constraint, not the beneficiary.

Never sell

A position should outlive the argument for buying it.

Add, never trim

When one holding gets too big, we fix it by adding to everything else, never by selling the winner.

Season law

A slice of every sweep goes to a reserve that only gets spent in a panic. Winters come, and the reserve is what lets us keep the never-sell rule when they do.

The toll booths rule is easiest to see from the ground up. At the bottom is the land, the water and the metal in it. Above that is power. Then the buildings that hold the computers, the fabs that make the chips, the chips themselves, the cloud, and at the top the language models. Every layer pays the one below it. We hold the bottom layers heavy and own the top through the index, because the top is where nobody can tell yet who wins.

The last piece is the Exploration in our name. Everything in that stack runs into the same three walls on the ground: land, power and permits. Some of it will go up to get around them, satellites and networks first, and maybe data centers after that. Whatever goes up pays for the ride first, so launch is the next toll booth. That conviction is the biggest position in the book. It’s also the one most likely to look wrong for years at a time, and we’re holding it anyway.

Around the care work is a family of companies that share one name, Culture. The Culture Group puts capital into care companies and helps run them. CultureAI is the software we install inside every one of them. The Culture Lab is where I write down what works, and what didn’t, in public. All of them run on the same four values: kindness, unity, humility and patience.

Now, what could go wrong. I could be wrong about what to own. Today the constraint in AI is power and chips. In thirty years it might be something nobody has a name for yet. If that happens the rule doesn’t change. We go find the new constraint and we add there.

The care itself could change too. I’ve worked in hospice, and hospice shows you life from the end. It’s where my conviction about healthy years comes from. If the work we fund helps people stay healthy longer, and that means they need companies like ours less, that’s ok.

Trust is the greatest economic force in the world. With it you need fewer rules, and without it you need more. The rules in this letter are for the money. The people run on trust, and on a promise I make to everyone who works with us: I don’t quit on you, and you don’t quit on me.

Anything worth anything takes a long time to build. So if you’re reading this in thirty years, check the count. If it still says zero, we kept our word.